Discrimination of lava dome activity styles using satellite-derived thermal structures

Abstract Thermal imaging from satellites is one methodology used for the monitoring and scientific investigation of volcanoes, including those characterised by active lava domes. To be most effective, the remote sensing techniques employed must allow the cause of any observed thermal anomaly to be identified, ideally using information contained within the remote sensing data itself, whilst using any ancillary field data to guide analysis and the necessary assumptions. This study investigates a method by which such discrimination maybe accomplished for activity at lava domes. For this purpose we use three Landsat Thematic Mapper (TM) scenes of Unzen Volcano (Japan) to determine the temperature structure of the active lava dome existing during 1991–1993. These data are of particularly high quality since the TM scenes were obtained during night-time overpasses and a large amount of supplementary data are available to parameterise the model used to retrieve the subpixel temperature structure of the dome surface. The results are matched to near-contemporaneous geological sketch maps of the then current situation in order to identify the TM pixel groupings corresponding to the different styles of dome activity, namely fumarolic degassing, exogenous and endogenous dome growth, and collapse deposits resulting from block and ash flows, small scale pyroclastic flows and rockfalls. The spatial and statistical characteristics of the resultant TM-derived temperature distributions are then investigated to determine rule-based criteria that may be used to differentiate the activity styles of lava domes based on their thermal structure within the remote sensing data. Results indicate that fumarolically active locations, regions of active dome growth, and areas of collapse deposition on the Unzen dome can be differentiated using only the statistical distribution of the hotspot temperatures and fractional areas identified via analysis of TM imagery. The statistics derived for hotspots at the fumarolically active areas show them to be, in general, significantly hotter (and smaller) than those found at the areas of active dome growth, whereas hotspots retrieved on the areas of collapse deposition are cooler (and larger). We find that the frequency distributions of hotspot temperatures extracted for the areas of endogenous and exogenous dome growth exhibit statistically inseparable means, but that discrimination between these two dome growth styles is possible using analysis of the spatial arrangement of the retrieved hotspots. Specifically the exogenous lobes are characterised by structures interpreted to be the effusing vent of new magma and the collapsing lobe front. These features are absent at locations showing only endogenous growth. It is hoped that the criteria developed here will prove useful during future quantitative analysis of the extended TM time-series available for Unzen, and for identification of thermal anomalies of uncertain origin at other active domes observed via satellite remote sensing.